arch/ia64/kernel/process.c

   1 /*
   2  * Architecture-specific setup.
   3  *
   4  * Copyright (C) 1998-2003 Hewlett-Packard Co
   5  *      David Mosberger-Tang <davidm@hpl.hp.com>
   6  * 04/11/17 Ashok Raj   <ashok.raj@intel.com> Added CPU Hotplug Support
   7  *
   8  * 2005-10-07 Keith Owens <kaos@sgi.com>
   9  *            Add notify_die() hooks.
  10  */
  11 #define __KERNEL_SYSCALLS__     /* see <asm/unistd.h> */
  12 #include <linux/config.h>
  13
  14 #include <linux/cpu.h>
  15 #include <linux/pm.h>
  16 #include <linux/elf.h>
  17 #include <linux/errno.h>
  18 #include <linux/kallsyms.h>
  19 #include <linux/kernel.h>
  20 #include <linux/mm.h>
  21 #include <linux/module.h>
  22 #include <linux/notifier.h>
  23 #include <linux/personality.h>
  24 #include <linux/sched.h>
  25 #include <linux/slab.h>
  26 #include <linux/smp_lock.h>
  27 #include <linux/stddef.h>
  28 #include <linux/thread_info.h>
  29 #include <linux/unistd.h>
  30 #include <linux/efi.h>
  31 #include <linux/interrupt.h>
  32 #include <linux/delay.h>
  33
  34 #include <asm/cpu.h>
  35 #include <asm/delay.h>
  36 #include <asm/elf.h>
  37 #include <asm/ia32.h>
  38 #include <asm/irq.h>
  39 #include <asm/kdebug.h>
  40 #include <asm/pgalloc.h>
  41 #include <asm/processor.h>
  42 #include <asm/sal.h>
  43 #include <asm/tlbflush.h>
  44 #include <asm/uaccess.h>
  45 #include <asm/unwind.h>
  46 #include <asm/user.h>
  47
  48 #include "entry.h"
  49
  50 #ifdef CONFIG_PERFMON
  51 # include <asm/perfmon.h>
  52 #endif
  53
  54 #include "sigframe.h"
  55
  56 void (*ia64_mark_idle)(int);
  57 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  58
  59 unsigned long boot_option_idle_override = 0;
  60 EXPORT_SYMBOL(boot_option_idle_override);
  61
  62 void
  63 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
  64 {
  65         unsigned long ip, sp, bsp;
  66         char buf[128];                  /* don't make it so big that it overflows the stack! */
  67
  68         printk("\nCall Trace:\n");
  69         do {
  70                 unw_get_ip(info, &ip);
  71                 if (ip == 0)
  72                         break;
  73
  74                 unw_get_sp(info, &sp);
  75                 unw_get_bsp(info, &bsp);
  76                 snprintf(buf, sizeof(buf),
  77                          " [<%016lx>] %%s\n"
  78                          "                                sp=%016lx bsp=%016lx\n",
  79                          ip, sp, bsp);
  80                 print_symbol(buf, ip);
  81         } while (unw_unwind(info) >= 0);
  82 }
  83
  84 void
  85 show_stack (struct task_struct *task, unsigned long *sp)
  86 {
  87         if (!task)
  88                 unw_init_running(ia64_do_show_stack, NULL);
  89         else {
  90                 struct unw_frame_info info;
  91
  92                 unw_init_from_blocked_task(&info, task);
  93                 ia64_do_show_stack(&info, NULL);
  94         }
  95 }
  96
  97 void
  98 dump_stack (void)
  99 {
 100         show_stack(NULL, NULL);
 101 }
 102
 103 EXPORT_SYMBOL(dump_stack);
 104
 105 void
 106 show_regs (struct pt_regs *regs)
 107 {
 108         unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
 109
 110         print_modules();
 111         printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
 112         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
 113                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
 114         print_symbol("ip is at %s\n", ip);
 115         printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
 116                regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
 117         printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
 118                regs->ar_rnat, regs->ar_bspstore, regs->pr);
 119         printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
 120                regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
 121         printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
 122         printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
 123         printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
 124                regs->f6.u.bits[1], regs->f6.u.bits[0],
 125                regs->f7.u.bits[1], regs->f7.u.bits[0]);
 126         printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
 127                regs->f8.u.bits[1], regs->f8.u.bits[0],
 128                regs->f9.u.bits[1], regs->f9.u.bits[0]);
 129         printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
 130                regs->f10.u.bits[1], regs->f10.u.bits[0],
 131                regs->f11.u.bits[1], regs->f11.u.bits[0]);
 132
 133         printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
 134         printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
 135         printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
 136         printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
 137         printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
 138         printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
 139         printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
 140         printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
 141         printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
 142
 143         if (user_mode(regs)) {
 144                 /* print the stacked registers */
 145                 unsigned long val, *bsp, ndirty;
 146                 int i, sof, is_nat = 0;
 147
 148                 sof = regs->cr_ifs & 0x7f;      /* size of frame */
 149                 ndirty = (regs->loadrs >> 19);
 150                 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
 151                 for (i = 0; i < sof; ++i) {
 152                         get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
 153                         printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
 154                                ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
 155                 }
 156         } else
 157                 show_stack(NULL, NULL);
 158 }
 159
 160 void
 161 do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
 162 {
 163         if (fsys_mode(current, &scr->pt)) {
 164                 /* defer signal-handling etc. until we return to privilege-level 0.  */
 165                 if (!ia64_psr(&scr->pt)->lp)
 166                         ia64_psr(&scr->pt)->lp = 1;
 167                 return;
 168         }
 169
 170 #ifdef CONFIG_PERFMON
 171         if (current->thread.pfm_needs_checking)
 172                 pfm_handle_work();
 173 #endif
 174
 175         /* deal with pending signal delivery */
 176         if (test_thread_flag(TIF_SIGPENDING))
 177                 ia64_do_signal(oldset, scr, in_syscall);
 178 }
 179
 180 static int pal_halt        = 1;
 181 static int can_do_pal_halt = 1;
 182
 183 static int __init nohalt_setup(char * str)
 184 {
 185         pal_halt = can_do_pal_halt = 0;
 186         return 1;
 187 }
 188 __setup("nohalt", nohalt_setup);
 189
 190 void
 191 update_pal_halt_status(int status)
 192 {
 193         can_do_pal_halt = pal_halt && status;
 194 }
 195
 196 /*
 197  * We use this if we don't have any better idle routine..
 198  */
 199 void
 200 default_idle (void)
 201 {
 202         local_irq_enable();
 203         while (!need_resched()) {
 204                 if (can_do_pal_halt)
 205                         safe_halt();
 206                 else
 207                         cpu_relax();
 208         }
 209 }
 210
 211 #ifdef CONFIG_HOTPLUG_CPU
 212 /* We don't actually take CPU down, just spin without interrupts. */
 213 static inline void play_dead(void)
 214 {
 215         extern void ia64_cpu_local_tick (void);
 216         unsigned int this_cpu = smp_processor_id();
 217
 218         /* Ack it */
 219         __get_cpu_var(cpu_state) = CPU_DEAD;
 220
 221         max_xtp();
 222         local_irq_disable();
 223         idle_task_exit();
 224         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 225         /*
 226          * The above is a point of no-return, the processor is
 227          * expected to be in SAL loop now.
 228          */
 229         BUG();
 230 }
 231 #else
 232 static inline void play_dead(void)
 233 {
 234         BUG();
 235 }
 236 #endif /* CONFIG_HOTPLUG_CPU */
 237
 238 void cpu_idle_wait(void)
 239 {
 240         unsigned int cpu, this_cpu = get_cpu();
 241         cpumask_t map;
 242
 243         set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
 244         put_cpu();
 245
 246         cpus_clear(map);
 247         for_each_online_cpu(cpu) {
 248                 per_cpu(cpu_idle_state, cpu) = 1;
 249                 cpu_set(cpu, map);
 250         }
 251
 252         __get_cpu_var(cpu_idle_state) = 0;
 253
 254         wmb();
 255         do {
 256                 ssleep(1);
 257                 for_each_online_cpu(cpu) {
 258                         if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
 259                                 cpu_clear(cpu, map);
 260                 }
 261                 cpus_and(map, map, cpu_online_map);
 262         } while (!cpus_empty(map));
 263 }
 264 EXPORT_SYMBOL_GPL(cpu_idle_wait);
 265
 266 void __attribute__((noreturn))
 267 cpu_idle (void)
 268 {
 269         void (*mark_idle)(int) = ia64_mark_idle;
 270         int cpu = smp_processor_id();
 271
 272         /* endless idle loop with no priority at all */
 273         while (1) {
 274                 if (can_do_pal_halt)
 275                         current_thread_info()->status &= ~TS_POLLING;
 276                 else
 277                         current_thread_info()->status |= TS_POLLING;
 278
 279                 if (!need_resched()) {
 280                         void (*idle)(void);
 281 #ifdef CONFIG_SMP
 282                         min_xtp();
 283 #endif
 284                         if (__get_cpu_var(cpu_idle_state))
 285                                 __get_cpu_var(cpu_idle_state) = 0;
 286
 287                         rmb();
 288                         if (mark_idle)
 289                                 (*mark_idle)(1);
 290
 291                         idle = pm_idle;
 292                         if (!idle)
 293                                 idle = default_idle;
 294                         (*idle)();
 295                         if (mark_idle)
 296                                 (*mark_idle)(0);
 297 #ifdef CONFIG_SMP
 298                         normal_xtp();
 299 #endif
 300                 }
 301                 preempt_enable_no_resched();
 302                 schedule();
 303                 preempt_disable();
 304                 check_pgt_cache();
 305                 if (cpu_is_offline(cpu))
 306                         play_dead();
 307         }
 308 }
 309
 310 void
 311 ia64_save_extra (struct task_struct *task)
 312 {
 313 #ifdef CONFIG_PERFMON
 314         unsigned long info;
 315 #endif
 316
 317         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 318                 ia64_save_debug_regs(&task->thread.dbr[0]);
 319
 320 #ifdef CONFIG_PERFMON
 321         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 322                 pfm_save_regs(task);
 323
 324         info = __get_cpu_var(pfm_syst_info);
 325         if (info & PFM_CPUINFO_SYST_WIDE)
 326                 pfm_syst_wide_update_task(task, info, 0);
 327 #endif
 328
 329 #ifdef CONFIG_IA32_SUPPORT
 330         if (IS_IA32_PROCESS(task_pt_regs(task)))
 331                 ia32_save_state(task);
 332 #endif
 333 }
 334
 335 void
 336 ia64_load_extra (struct task_struct *task)
 337 {
 338 #ifdef CONFIG_PERFMON
 339         unsigned long info;
 340 #endif
 341
 342         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 343                 ia64_load_debug_regs(&task->thread.dbr[0]);
 344
 345 #ifdef CONFIG_PERFMON
 346         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 347                 pfm_load_regs(task);
 348
 349         info = __get_cpu_var(pfm_syst_info);
 350         if (info & PFM_CPUINFO_SYST_WIDE)
 351                 pfm_syst_wide_update_task(task, info, 1);
 352 #endif
 353
 354 #ifdef CONFIG_IA32_SUPPORT
 355         if (IS_IA32_PROCESS(task_pt_regs(task)))
 356                 ia32_load_state(task);
 357 #endif
 358 }
 359
 360 /*
 361  * Copy the state of an ia-64 thread.
 362  *
 363  * We get here through the following  call chain:
 364  *
 365  *      from user-level:        from kernel:
 366  *
 367  *      <clone syscall>         <some kernel call frames>
 368  *      sys_clone                  :
 369  *      do_fork                 do_fork
 370  *      copy_thread             copy_thread
 371  *
 372  * This means that the stack layout is as follows:
 373  *
 374  *      +---------------------+ (highest addr)
 375  *      |   struct pt_regs    |
 376  *      +---------------------+
 377  *      | struct switch_stack |
 378  *      +---------------------+
 379  *      |                     |
 380  *      |    memory stack     |
 381  *      |                     | <-- sp (lowest addr)
 382  *      +---------------------+
 383  *
 384  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 385  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 386  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 387  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 388  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 389  * so there is nothing to worry about.
 390  */
 391 int
 392 copy_thread (int nr, unsigned long clone_flags,
 393              unsigned long user_stack_base, unsigned long user_stack_size,
 394              struct task_struct *p, struct pt_regs *regs)
 395 {
 396         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 397         struct switch_stack *child_stack, *stack;
 398         unsigned long rbs, child_rbs, rbs_size;
 399         struct pt_regs *child_ptregs;
 400         int retval = 0;
 401
 402 #ifdef CONFIG_SMP
 403         /*
 404          * For SMP idle threads, fork_by_hand() calls do_fork with
 405          * NULL regs.
 406          */
 407         if (!regs)
 408                 return 0;
 409 #endif
 410
 411         stack = ((struct switch_stack *) regs) - 1;
 412
 413         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 414         child_stack = (struct switch_stack *) child_ptregs - 1;
 415
 416         /* copy parent's switch_stack & pt_regs to child: */
 417         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 418
 419         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 420         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 421         rbs_size = stack->ar_bspstore - rbs;
 422
 423         /* copy the parent's register backing store to the child: */
 424         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 425
 426         if (likely(user_mode(child_ptregs))) {
 427                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 428                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 429                 if (user_stack_base) {
 430                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 431                         child_ptregs->ar_bspstore = user_stack_base;
 432                         child_ptregs->ar_rnat = 0;
 433                         child_ptregs->loadrs = 0;
 434                 }
 435         } else {
 436                 /*
 437                  * Note: we simply preserve the relative position of
 438                  * the stack pointer here.  There is no need to
 439                  * allocate a scratch area here, since that will have
 440                  * been taken care of by the caller of sys_clone()
 441                  * already.
 442                  */
 443                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 444                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 445         }
 446         child_stack->ar_bspstore = child_rbs + rbs_size;
 447         if (IS_IA32_PROCESS(regs))
 448                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 449         else
 450                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 451
 452         /* copy parts of thread_struct: */
 453         p->thread.ksp = (unsigned long) child_stack - 16;
 454
 455         /* stop some PSR bits from being inherited.
 456          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 457          * therefore we must specify them explicitly here and not include them in
 458          * IA64_PSR_BITS_TO_CLEAR.
 459          */
 460         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 461                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 462
 463         /*
 464          * NOTE: The calling convention considers all floating point
 465          * registers in the high partition (fph) to be scratch.  Since
 466          * the only way to get to this point is through a system call,
 467          * we know that the values in fph are all dead.  Hence, there
 468          * is no need to inherit the fph state from the parent to the
 469          * child and all we have to do is to make sure that
 470          * IA64_THREAD_FPH_VALID is cleared in the child.
 471          *
 472          * XXX We could push this optimization a bit further by
 473          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 474          * However, it's not clear this is worth doing.  Also, it
 475          * would be a slight deviation from the normal Linux system
 476          * call behavior where scratch registers are preserved across
 477          * system calls (unless used by the system call itself).
 478          */
 479 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 480                                          | IA64_THREAD_PM_VALID)
 481 #       define THREAD_FLAGS_TO_SET      0
 482         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 483                            | THREAD_FLAGS_TO_SET);
 484         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 485 #ifdef CONFIG_IA32_SUPPORT
 486         /*
 487          * If we're cloning an IA32 task then save the IA32 extra
 488          * state from the current task to the new task
 489          */
 490         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 491                 ia32_save_state(p);
 492                 if (clone_flags & CLONE_SETTLS)
 493                         retval = ia32_clone_tls(p, child_ptregs);
 494
 495                 /* Copy partially mapped page list */
 496                 if (!retval)
 497                         retval = ia32_copy_partial_page_list(p, clone_flags);
 498         }
 499 #endif
 500
 501 #ifdef CONFIG_PERFMON
 502         if (current->thread.pfm_context)
 503                 pfm_inherit(p, child_ptregs);
 504 #endif
 505         return retval;
 506 }
 507
 508 static void
 509 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 510 {
 511         unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
 512         elf_greg_t *dst = arg;
 513         struct pt_regs *pt;
 514         char nat;
 515         int i;
 516
 517         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 518
 519         if (unw_unwind_to_user(info) < 0)
 520                 return;
 521
 522         unw_get_sp(info, &sp);
 523         pt = (struct pt_regs *) (sp + 16);
 524
 525         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 526
 527         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 528                 return;
 529
 530         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 531                   &ar_rnat);
 532
 533         /*
 534          * coredump format:
 535          *      r0-r31
 536          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 537          *      predicate registers (p0-p63)
 538          *      b0-b7
 539          *      ip cfm user-mask
 540          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 541          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 542          */
 543
 544         /* r0 is zero */
 545         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 546                 unw_get_gr(info, i, &dst[i], &nat);
 547                 if (nat)
 548                         nat_bits |= mask;
 549                 mask <<= 1;
 550         }
 551         dst[32] = nat_bits;
 552         unw_get_pr(info, &dst[33]);
 553
 554         for (i = 0; i < 8; ++i)
 555                 unw_get_br(info, i, &dst[34 + i]);
 556
 557         unw_get_rp(info, &ip);
 558         dst[42] = ip + ia64_psr(pt)->ri;
 559         dst[43] = cfm;
 560         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 561
 562         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 563         /*
 564          * For bsp and bspstore, unw_get_ar() would return the kernel
 565          * addresses, but we need the user-level addresses instead:
 566          */
 567         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 568         dst[47] = pt->ar_bspstore;
 569         dst[48] = ar_rnat;
 570         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 571         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 572         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 573         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 574         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 575         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 576         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 577         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 578 }
 579
 580 void
 581 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 582 {
 583         elf_fpreg_t *dst = arg;
 584         int i;
 585
 586         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 587
 588         if (unw_unwind_to_user(info) < 0)
 589                 return;
 590
 591         /* f0 is 0.0, f1 is 1.0 */
 592
 593         for (i = 2; i < 32; ++i)
 594                 unw_get_fr(info, i, dst + i);
 595
 596         ia64_flush_fph(task);
 597         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 598                 memcpy(dst + 32, task->thread.fph, 96*16);
 599 }
 600
 601 void
 602 do_copy_regs (struct unw_frame_info *info, void *arg)
 603 {
 604         do_copy_task_regs(current, info, arg);
 605 }
 606
 607 void
 608 do_dump_fpu (struct unw_frame_info *info, void *arg)
 609 {
 610         do_dump_task_fpu(current, info, arg);
 611 }
 612
 613 int
 614 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
 615 {
 616         struct unw_frame_info tcore_info;
 617
 618         if (current == task) {
 619                 unw_init_running(do_copy_regs, regs);
 620         } else {
 621                 memset(&tcore_info, 0, sizeof(tcore_info));
 622                 unw_init_from_blocked_task(&tcore_info, task);
 623                 do_copy_task_regs(task, &tcore_info, regs);
 624         }
 625         return 1;
 626 }
 627
 628 void
 629 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 630 {
 631         unw_init_running(do_copy_regs, dst);
 632 }
 633
 634 int
 635 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
 636 {
 637         struct unw_frame_info tcore_info;
 638
 639         if (current == task) {
 640                 unw_init_running(do_dump_fpu, dst);
 641         } else {
 642                 memset(&tcore_info, 0, sizeof(tcore_info));
 643                 unw_init_from_blocked_task(&tcore_info, task);
 644                 do_dump_task_fpu(task, &tcore_info, dst);
 645         }
 646         return 1;
 647 }
 648
 649 int
 650 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 651 {
 652         unw_init_running(do_dump_fpu, dst);
 653         return 1;       /* f0-f31 are always valid so we always return 1 */
 654 }
 655
 656 long
 657 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 658             struct pt_regs *regs)
 659 {
 660         char *fname;
 661         int error;
 662
 663         fname = getname(filename);
 664         error = PTR_ERR(fname);
 665         if (IS_ERR(fname))
 666                 goto out;
 667         error = do_execve(fname, argv, envp, regs);
 668         putname(fname);
 669 out:
 670         return error;
 671 }
 672
 673 pid_t
 674 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 675 {
 676         extern void start_kernel_thread (void);
 677         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 678         struct {
 679                 struct switch_stack sw;
 680                 struct pt_regs pt;
 681         } regs;
 682
 683         memset(&regs, 0, sizeof(regs));
 684         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 685         regs.pt.r1 = helper_fptr[1];            /* set GP */
 686         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 687         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 688         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 689         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 690         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 691         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 692         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 693         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 694         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 695 }
 696 EXPORT_SYMBOL(kernel_thread);
 697
 698 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 699 int
 700 kernel_thread_helper (int (*fn)(void *), void *arg)
 701 {
 702 #ifdef CONFIG_IA32_SUPPORT
 703         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 704                 /* A kernel thread is always a 64-bit process. */
 705                 current->thread.map_base  = DEFAULT_MAP_BASE;
 706                 current->thread.task_size = DEFAULT_TASK_SIZE;
 707                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 708                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 709         }
 710 #endif
 711         return (*fn)(arg);
 712 }
 713
 714 /*
 715  * Flush thread state.  This is called when a thread does an execve().
 716  */
 717 void
 718 flush_thread (void)
 719 {
 720         /* drop floating-point and debug-register state if it exists: */
 721         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 722         ia64_drop_fpu(current);
 723 #ifdef CONFIG_IA32_SUPPORT
 724         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 725                 ia32_drop_partial_page_list(current);
 726                 current->thread.task_size = IA32_PAGE_OFFSET;
 727                 set_fs(USER_DS);
 728         }
 729 #endif
 730 }
 731
 732 /*
 733  * Clean up state associated with current thread.  This is called when
 734  * the thread calls exit().
 735  */
 736 void
 737 exit_thread (void)
 738 {
 739
 740         ia64_drop_fpu(current);
 741 #ifdef CONFIG_PERFMON
 742        /* if needed, stop monitoring and flush state to perfmon context */
 743         if (current->thread.pfm_context)
 744                 pfm_exit_thread(current);
 745
 746         /* free debug register resources */
 747         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 748                 pfm_release_debug_registers(current);
 749 #endif
 750         if (IS_IA32_PROCESS(task_pt_regs(current)))
 751                 ia32_drop_partial_page_list(current);
 752 }
 753
 754 unsigned long
 755 get_wchan (struct task_struct *p)
 756 {
 757         struct unw_frame_info info;
 758         unsigned long ip;
 759         int count = 0;
 760
 761         /*
 762          * Note: p may not be a blocked task (it could be current or
 763          * another process running on some other CPU.  Rather than
 764          * trying to determine if p is really blocked, we just assume
 765          * it's blocked and rely on the unwind routines to fail
 766          * gracefully if the process wasn't really blocked after all.
 767          * --davidm 99/12/15
 768          */
 769         unw_init_from_blocked_task(&info, p);
 770         do {
 771                 if (unw_unwind(&info) < 0)
 772                         return 0;
 773                 unw_get_ip(&info, &ip);
 774                 if (!in_sched_functions(ip))
 775                         return ip;
 776         } while (count++ < 16);
 777         return 0;
 778 }
 779
 780 void
 781 cpu_halt (void)
 782 {
 783         pal_power_mgmt_info_u_t power_info[8];
 784         unsigned long min_power;
 785         int i, min_power_state;
 786
 787         if (ia64_pal_halt_info(power_info) != 0)
 788                 return;
 789
 790         min_power_state = 0;
 791         min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
 792         for (i = 1; i < 8; ++i)
 793                 if (power_info[i].pal_power_mgmt_info_s.im
 794                     && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
 795                         min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
 796                         min_power_state = i;
 797                 }
 798
 799         while (1)
 800                 ia64_pal_halt(min_power_state);
 801 }
 802
 803 void
 804 machine_restart (char *restart_cmd)
 805 {
 806         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 807         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 808 }
 809
 810 void
 811 machine_halt (void)
 812 {
 813         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 814         cpu_halt();
 815 }
 816
 817 void
 818 machine_power_off (void)
 819 {
 820         if (pm_power_off)
 821                 pm_power_off();
 822         machine_halt();
 823 }
 824